Sampling methods

For detailed information, see http://db.cger.nies.go.jp/JaLTER/ER_DataPapers/archives/2017/ERDP-2017-05/metadata A digital voice recorder (DS-850, Olympus, Tokyo, Japan; Fig. 2) was installed at each monitoring site during May–July in each year. The recorders were adjusted to timer-recording mode and recorded for 10 min before and after sunrise (total 20 min) every day until the batteries were depleted. The recorders were fixed to tripod stands at a height of about 0.9 m. The recorded data were split into 1-min segments in MP3 (124 kbps) format, which was treated as the minimum sample unit. We identified species of birds from acoustic data. Bioacoustics signals is a promised source of information for avian species identification (Lopes et al. 2011), and acoustic monitoring can produce similar result as traditional on-site survey methods in comparative ecological studies (Haselmayer & Quinn 2000; Hobson et al. 2002; Klingbeil & Willig 2015). Because the number of segments was very large, we chose a subset of segments evenly throughout the sampling period (8.08 days/site and 17.2 segments/day/site in 2014, 7.89 days/site and 17.8 segments/day/site in 2015, 5.74 days/site and 17.1 segments/day/site in 2016, 4.28 days/site and 16.8 segments/day/site in 2017, 3.75 days/site and 20 segments/day/site in 2018, 3.85 days/site and 20 segments/day/site in 2019 and 　3.9 days/site and 20 segments/day/site in 2020, in average). A total of 7,222 of the 45,540 segments were chosen in 2014, 8,017 of the 46,440 segments were chosen in 2015, 5,289 of the 42,440 segments were chosen in 2016, 4,092 of the 46,680 segments were chosen in 2017, 4,200 of the 43,580 segments were chosen in 2018, 4,000 of the 42,720 segments were chosen in 2019 and 3,900 of the 44,759 segments were chosen in 2020. Species that appeared in each segment were identified by experts and their presence-absence was recorded. Some of the segments were identified through a citizen-scientific project, “Bird Data Challenge (Fukasawa et al. 2017)”, in which we listened to audio data and prepared a species list of birds with involvement of local citizen experts. Data identified through the Bird Data Challenge was checked by authors or other experts to correct misspecifications of species. We were careful to avoid the negative impact (e.g., pressure of photographing and illegal capture) on endangered and attractive species when we made the species distribution data accessible online. We left the location ID blank and assigned the mean latitude and longitude of the study area in the presence-absence records for endangered species (i.e., species ranked VU, EN, and CR in the National or Prefectural Red List) and species attracting particular public interest (Terpsiphone atrocaudata and Halcyon coromanda).